Ultra-Efficient Photogrammetry with Pix4Dmapper Pro’s Multi-Camera Rig Processing
As a professional photogrammetry software, Pix4Dmapper Pro can easily handle multiple cameras in one project processing or for project merging. We commonly recommend our users to execute grid flights with nadir images for digital surface models (DSMs) and ortho mosaics, and double-grid or circular flights with oblique images for 3D modeling. Now we would also like to suggest another way for reconstructing the object space efficiently – Taking images with multiple cameras at the same time, which will give you the best results with the shortest acquisition and processing time
What are multi-camera rigs?
The systems of multiple cameras are so-called “multi-camera rigs. They are systems which connect several cameras by rigs. With one trigger command, the systems enable all cameras to take images simultaneously.
By “simultaneously,” we refer to a time difference much less than the image read-out time, approximated to be around 10 milliseconds. In Pix4Dmapper Pro, having parameters that describe how the cameras are connected are crucial for calibrating multi-camera rigs successfully. Those parameters include the spatial relation among the cameras: the difference in translations and rotations along the x, y, z axes, are called the rig relatives.
There are quite a few multi-camera rigs on the market, such as XCAM Ultra and RGB-Thermal from WaldoAir, MIDAS from Track’Air, and RCD 30 from Leica. These are all examples of well-calibrated camera systems suitable for photogrammetric applications.
Why do we use multi-camera rigs?
1. Maximum coverage, minimum time
Photogrammetry techniques generate results based purely on image content, and a general rule is more image coverage for better results. Take 3D urban modeling as an example: the more coverage on building façade surfaces from all directions, the sharper edges your final 3D model will have. In order to get more image content of the facades, you could use a fisheye camera for wider coverage, capture images with different camera angles in several flights, or use multi-camera rigs.
With the same acquisition plans, fisheye cameras will give you wider coverage than perspective cameras. However, there will be ununiform ground sampling distance (GSD) in the output even though Pix4Dmapper Pro is capable of correcting the lens distortions. One of the advantages of using multi-camera rigs for image acquisition is time saved. Imagine that instead of flying over the same area nine times with nine different camera angle settings, people would only need to fly once with a nine-camera rigs.
Calibrated image positions and a small part of the 3D model from oblique MIDAS project in Pix4Dmapper Pro’s rayCloud
2. Easier calibration of more-challenging sensors
Multi-camera rigs are also practical when certain other cameras are not as easy to calibrate: Thermal cameras for example. People who map using thermal cameras may have found that they capture images of much smaller extent. In the current market, consumer-grade cameras give you images of 3000 x 4000 pixels and up, while a popularly-used thermal camera only captures images at 640 x 512 pixels. As a result, less feature points can be found in each image and thus much fewer matches among images. In such cases, users will need to either increase the image overlap or sacrifice the details (lower ground sampling distance) in order to get workable results.
However, it is now able to better calibrate a thermal camera by connecting it with higher-resolution RGB cameras. Once the rig relatives – the translation and orientation among the cameras – are known or computed, Pix4D software is able to calibrate the thermal camera using the features found in the higher-resolution RGB camera images. As a plus, you get both RGB and thermal mapping outputs of the same region from one data acquisition!
Calibrated image positions and the entire WaldoAir thermal-RGB project in Pix4Dmapper Pro’s rayCloud
3. Customizing your own multi-camera rig system for optimal usage
More and more people are realizing the convenience of acquiring images using rigged cameras. They can be put on any platform, as well as being executed stand-alone for terrestrial or close-range projects.
Zoom of calibrated image positions from a custom Canon camera rigs in Pix4Dmapper Pro’s rayCloud
Building your own multi-camera rigs is a good idea because you can choose the cameras and assemble them in a way most suitable for your mapping purposes. The example custom rigs is composed of six cameras where the middle two are shooting 20 degrees toward each other to capture close-nadir image content, and the four surrounded cameras set to capture façade information
How does Pix4Dmapper Pro handle multi-camera rigs?
We have focused a lot on how efficient and convenient it is to acquire images using multi-camera rigs, and now we finally come to the point – How can camera rig users benefit from Pix4Dmapper Pro?
A. It takes images folder-by-folder and creates a rig system for all detected cameras
Images acquired with the same camera have to be grouped in one folder. Instead of importing all images, users need to import them folder-based in order for the software to know they are from multi-camera rigs.
Import multi-camera rig images by folders and select “new project with camera rigs”
B. It uses the given relatives to optimize multi-camera rigs in the most efficient way
Getting initial values from the rig calibration report is always a good thing, although it might not always be available. Here we can see the calibrated translations among the rig cameras, which the users can input directly into Pix4Dmapper Pro using its rig interface.
Lab-calibrated relatives of octoliqbue MIDAS camera
Translations among the cameras are normally easier to measure than the rotations, and they must be given to the software. In some cases the rotations are almost impossible to know prior to the calibration process, so users could temporarily leave them blank and let the software to find the precise rotation during the image matching process. Based on the reliability of the initial rig relatives, users can choose to compute all or only certain parts of the rig relatives.
Pix4Dmapper Pro’s multi-camera rig processing interface
Once the multi-camera rigs have been calibrated for the first time, the relatives will be saved in Pix4Dmapper Pro’s database. Then, images taken by these multi-camera rigs will be automatically detected for the next time. This will minimize processing time. Take a nine-rig image dataset for example, processing all the images with known rigs would only require one-ninth of the time that processing them all separately would. Here are some options users can select according to the reliability of the rig relatives:
Options |
Action |
Rely on Relatives |
Relatives MUST be precisely known. Fix previous calibrated translations and orientations without having the software optimize them based on project images |
Optimize Relative Rotation |
Fix previous calibrated translations and have the software optimize only the rotations based on all project images |
Optimize Relative Rotationwith a subset of secondaries |
Fix previous calibrated translations and have the software optimize only the rotations based on a subset of project images |
Optimize Independently |
Compute the relatives as the cameras are not perfectly rigged |
Disable rig |
Ignore all rig constraints and calibrate as if there was no rig |
Calibrate rig |
Calibrate all cameras independently, estimate rig relatives, and use these for all cameras. Requires GCPs for the scale |
C. It does NOT consider unsynchronized images as multi-camera rigs
Rig processing is only suitable for images acquired simultaneously. Under such conditions, the images are following the constraint of those rig relatives.
In the left figure, we see that both cameras from the rig are capturing images at the same time, and this allows the image pairs from the same multi-camera rig to always keep the same translations and rotations.
Now, what if the triggering of rig cameras is not synchronized? It is especially a problem when the cameras and object have relative movement. In the right figure, we can see that the two cameras capture images at different moments while the aircraft is moving.
This breaks the constraint of the rig relatives and risks to fail image calibration. In this case, the cameras would need to be optimized individually and the images need to be processed separately in Pix4Dmapper Pro using the standard workflow instead of multi-camera rig processing.
